You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
 
 
 
 
 
 

411 lines
12 KiB

// Upgrade NOTE: replaced 'mul(UNITY_MATRIX_MVP,*)' with 'UnityObjectToClipPos(*)'
// Upgrade NOTE: commented out 'float4x4 _WorldToCamera', a built-in variable
// Upgrade NOTE: replaced '_WorldToCamera' with 'unity_WorldToCamera'
// --------
// Additional options for further customization
// --------
// By default, a fixed sampling pattern is used in the AO estimator.
// Although this gives preferable results in most cases, a completely
// random sampling pattern could give aesthetically good results in some
// cases. Comment out the line below to use the random pattern instead of
// the fixed one.
#define FIXED_SAMPLING_PATTERN 1
// The constant below determines the contrast of occlusion. Altough this
// allows intentional over/under occlusion, currently is not exposed to the
// editor, because it’s thought to be rarely useful.
static const float kContrast = 0.6;
// The constant below controls the geometry-awareness of the blur filter.
// The higher value, the more sensitive it is.
static const float kGeom = 50;
// The constants below are used in the AO estimator. Beta is mainly used
// for suppressing self-shadowing noise, and Epsilon is used to prevent
// calculation underflow. See the paper (Morgan 2011 http://goo.gl/2iz3P)
// for further details of these constants.
static const float kBeta = 0.002;
static const float kEpsilon = 1e-4;
// --------
#include "UnityCG.cginc"
// Global shader properties
sampler2D _CameraGBufferTexture2;
sampler2D_float _CameraDepthTexture;
sampler2D _CameraDepthNormalsTexture;
// float4x4 _WorldToCamera;
// Sample count
// Use a constant on GLES2 (basically it doesn't support dynamic looping).
#if SHADER_API_GLES
static const int _SampleCount = 5;
#else
int _SampleCount;
#endif
sampler2D _MainTex;
float4 _MainTex_TexelSize;
sampler2D _OcclusionTexture;
// Material shader properties
half _Intensity;
float _Radius;
float _TargetScale;
float2 _BlurVector;
// Utility for sin/cos
float2 CosSin(float theta)
{
float sn, cs;
sincos(theta, sn, cs);
return float2(cs, sn);
}
// Gamma encoding function for AO value
// (do nothing if in the linear mode)
half EncodeAO(half x)
{
// Gamma encoding
half x_g = 1 - pow(1 - x, 1 / 2.2);
// ColorSpaceLuminance.w is 0 (gamma) or 1 (linear).
return lerp(x_g, x, unity_ColorSpaceLuminance.w);
}
// Pseudo random number generator with 2D argument
float UVRandom(float u, float v)
{
float f = dot(float2(12.9898, 78.233), float2(u, v));
return frac(43758.5453 * sin(f));
}
// Interleaved gradient function from Jimenez 2014 http://goo.gl/eomGso
float GradientNoise(float2 uv)
{
uv = floor(uv * _ScreenParams.xy);
float f = dot(float2(0.06711056f, 0.00583715f), uv);
return frac(52.9829189f * frac(f));
}
// Boundary check for depth sampler
// (returns a very large value if it lies out of bounds)
float CheckBounds(float2 uv, float d)
{
float ob = any(uv < 0) + any(uv > 1);
#if defined(UNITY_REVERSED_Z)
ob += (d <= 0.00001);
#else
ob += (d >= 0.99999);
#endif
return ob * 1e8;
}
// Depth/normal sampling functions
float SampleDepth(float2 uv)
{
#if SOURCE_GBUFFER || SOURCE_DEPTH
float d = SAMPLE_DEPTH_TEXTURE(_CameraDepthTexture, uv);
return LinearEyeDepth(d) + CheckBounds(uv, d);
#else
float4 cdn = tex2D(_CameraDepthNormalsTexture, uv);
float d = DecodeFloatRG(cdn.zw);
return d * _ProjectionParams.z + CheckBounds(uv, d);
#endif
}
float3 SampleNormal(float2 uv)
{
#if SOURCE_GBUFFER
float3 norm = tex2D(_CameraGBufferTexture2, uv).xyz;
norm = norm * 2 - any(norm); // gets (0,0,0) when norm == 0
return mul((float3x3)unity_WorldToCamera, norm);
#else
float4 cdn = tex2D(_CameraDepthNormalsTexture, uv);
return DecodeViewNormalStereo(cdn) * float3(1, 1, -1);
#endif
}
float SampleDepthNormal(float2 uv, out float3 normal)
{
#if SOURCE_GBUFFER || SOURCE_DEPTH
normal = SampleNormal(uv);
return SampleDepth(uv);
#else
float4 cdn = tex2D(_CameraDepthNormalsTexture, uv);
normal = DecodeViewNormalStereo(cdn) * float3(1, 1, -1);
float d = DecodeFloatRG(cdn.zw);
return d * _ProjectionParams.z + CheckBounds(uv, d);
#endif
}
// Reconstruct view-space position from UV and depth.
// p11_22 = (unity_CameraProjection._11, unity_CameraProjection._22)
// p13_31 = (unity_CameraProjection._13, unity_CameraProjection._23)
float3 ReconstructViewPos(float2 uv, float depth, float2 p11_22, float2 p13_31)
{
return float3((uv * 2 - 1 - p13_31) / p11_22, 1) * depth;
}
// Normal vector comparer (for geometry-aware weighting)
half CompareNormal(half3 d1, half3 d2)
{
return pow((dot(d1, d2) + 1) * 0.5, kGeom);
}
// Final combiner function
half3 CombineOcclusion(half3 src, half3 ao)
{
return lerp(src, 0, EncodeAO(ao));
}
// Sample point picker
float3 PickSamplePoint(float2 uv, float index)
{
// Uniformaly distributed points on a unit sphere http://goo.gl/X2F1Ho
#if FIXED_SAMPLING_PATTERN
float gn = GradientNoise(uv * _TargetScale);
float u = frac(UVRandom(0, index) + gn) * 2 - 1;
float theta = (UVRandom(1, index) + gn) * UNITY_PI * 2;
#else
float u = UVRandom(uv.x + _Time.x, uv.y + index) * 2 - 1;
float theta = UVRandom(-uv.x - _Time.x, uv.y + index) * UNITY_PI * 2;
#endif
float3 v = float3(CosSin(theta) * sqrt(1 - u * u), u);
// Make them distributed between [0, _Radius]
float l = sqrt((index + 1) / _SampleCount) * _Radius;
return v * l;
}
// Occlusion estimator function
float EstimateOcclusion(float2 uv)
{
// Parameters used in coordinate conversion
float3x3 proj = (float3x3)unity_CameraProjection;
float2 p11_22 = float2(unity_CameraProjection._11, unity_CameraProjection._22);
float2 p13_31 = float2(unity_CameraProjection._13, unity_CameraProjection._23);
// View space normal and depth
float3 norm_o;
float depth_o = SampleDepthNormal(uv, norm_o);
#if SOURCE_DEPTHNORMALS
// Offset the depth value to avoid precision error.
// (depth in the DepthNormals mode has only 16-bit precision)
depth_o -= _ProjectionParams.z / 65536;
#endif
// Reconstruct the view-space position.
float3 vpos_o = ReconstructViewPos(uv, depth_o, p11_22, p13_31);
// Distance-based AO estimator based on Morgan 2011 http://goo.gl/2iz3P
float ao = 0.0;
for (int s = 0; s < _SampleCount; s++)
{
// Sample point
#if SHADER_API_D3D11
// This 'floor(1.0001 * s)' operation is needed to avoid a NVidia
// shader issue. This issue is only observed on DX11.
float3 v_s1 = PickSamplePoint(uv, floor(1.0001 * s));
#else
float3 v_s1 = PickSamplePoint(uv, s);
#endif
v_s1 = faceforward(v_s1, -norm_o, v_s1);
float3 vpos_s1 = vpos_o + v_s1;
// Reproject the sample point
float3 spos_s1 = mul(proj, vpos_s1);
float2 uv_s1 = (spos_s1.xy / vpos_s1.z + 1) * 0.5;
// Depth at the sample point
float depth_s1 = SampleDepth(uv_s1);
// Relative position of the sample point
float3 vpos_s2 = ReconstructViewPos(uv_s1, depth_s1, p11_22, p13_31);
float3 v_s2 = vpos_s2 - vpos_o;
// Estimate the obscurance value
float a1 = max(dot(v_s2, norm_o) - kBeta * depth_o, 0);
float a2 = dot(v_s2, v_s2) + kEpsilon;
ao += a1 / a2;
}
ao *= _Radius; // intensity normalization
// Apply other parameters.
return pow(ao * _Intensity / _SampleCount, kContrast);
}
// Geometry-aware separable blur filter (large kernel)
half SeparableBlurLarge(sampler2D tex, float2 uv, float2 delta)
{
#if !SHADER_API_MOBILE
// 9-tap Gaussian blur with adaptive sampling
float2 uv1a = uv - delta;
float2 uv1b = uv + delta;
float2 uv2a = uv - delta * 2;
float2 uv2b = uv + delta * 2;
float2 uv3a = uv - delta * 3.2307692308;
float2 uv3b = uv + delta * 3.2307692308;
half3 n0 = SampleNormal(uv);
half w0 = 0.37004405286;
half w1a = CompareNormal(n0, SampleNormal(uv1a)) * 0.31718061674;
half w1b = CompareNormal(n0, SampleNormal(uv1b)) * 0.31718061674;
half w2a = CompareNormal(n0, SampleNormal(uv2a)) * 0.19823788546;
half w2b = CompareNormal(n0, SampleNormal(uv2b)) * 0.19823788546;
half w3a = CompareNormal(n0, SampleNormal(uv3a)) * 0.11453744493;
half w3b = CompareNormal(n0, SampleNormal(uv3b)) * 0.11453744493;
half s = tex2D(_MainTex, uv).r * w0;
s += tex2D(_MainTex, uv1a).r * w1a;
s += tex2D(_MainTex, uv1b).r * w1b;
s += tex2D(_MainTex, uv2a).r * w2a;
s += tex2D(_MainTex, uv2b).r * w2b;
s += tex2D(_MainTex, uv3a).r * w3a;
s += tex2D(_MainTex, uv3b).r * w3b;
return s / (w0 + w1a + w1b + w2a + w2b + w3a + w3b);
#else
// 9-tap Gaussian blur with linear sampling
// (less quality but slightly fast)
float2 uv1a = uv - delta * 1.3846153846;
float2 uv1b = uv + delta * 1.3846153846;
float2 uv2a = uv - delta * 3.2307692308;
float2 uv2b = uv + delta * 3.2307692308;
half3 n0 = SampleNormal(uv);
half w0 = 0.2270270270;
half w1a = CompareNormal(n0, SampleNormal(uv1a)) * 0.3162162162;
half w1b = CompareNormal(n0, SampleNormal(uv1b)) * 0.3162162162;
half w2a = CompareNormal(n0, SampleNormal(uv2a)) * 0.0702702703;
half w2b = CompareNormal(n0, SampleNormal(uv2b)) * 0.0702702703;
half s = tex2D(_MainTex, uv).r * w0;
s += tex2D(_MainTex, uv1a).r * w1a;
s += tex2D(_MainTex, uv1b).r * w1b;
s += tex2D(_MainTex, uv2a).r * w2a;
s += tex2D(_MainTex, uv2b).r * w2b;
return s / (w0 + w1a + w1b + w2a + w2b);
#endif
}
// Geometry-aware separable blur filter (small kernel)
half SeparableBlurSmall(sampler2D tex, float2 uv, float2 delta)
{
float2 uv1 = uv - delta;
float2 uv2 = uv + delta;
half3 n0 = SampleNormal(uv);
half w0 = 2;
half w1 = CompareNormal(n0, SampleNormal(uv1));
half w2 = CompareNormal(n0, SampleNormal(uv2));
half s = tex2D(_MainTex, uv).r * w0;
s += tex2D(_MainTex, uv1).r * w1;
s += tex2D(_MainTex, uv2).r * w2;
return s / (w0 + w1 + w2);
}
// Occlusion estimation pass
half4 frag_ao(v2f_img i) : SV_Target
{
return EstimateOcclusion(i.uv);
}
// Noise reduction pass (large kernel)
half4 frag_blur1(v2f_img i) : SV_Target
{
float2 delta = _MainTex_TexelSize.xy * _BlurVector;
return SeparableBlurLarge(_MainTex, i.uv, delta);
}
// Noise reduction pass (small kernel)
half4 frag_blur2(v2f_img i) : SV_Target
{
float2 delta = _MainTex_TexelSize.xy * _BlurVector;
return SeparableBlurSmall(_MainTex, i.uv, delta);
}
// Combiner pass for the forward mode
struct v2f_multitex
{
float4 pos : SV_POSITION;
float2 uv0 : TEXCOORD0;
float2 uv1 : TEXCOORD1;
};
v2f_multitex vert_multitex(appdata_img v)
{
// Handles vertically-flipped case.
float vflip = sign(_MainTex_TexelSize.y);
v2f_multitex o;
o.pos = UnityObjectToClipPos(v.vertex);
o.uv0 = v.texcoord.xy;
o.uv1 = (v.texcoord.xy - 0.5) * float2(1, vflip) + 0.5;
return o;
}
half4 frag_combine(v2f_multitex i) : SV_Target
{
half4 src = tex2D(_MainTex, i.uv0);
half ao = tex2D(_OcclusionTexture, i.uv1).r;
return half4(CombineOcclusion(src.rgb, ao), src.a);
}
// Combiner pass for the ambient-only mode
v2f_img vert_gbuffer(appdata_img v)
{
v2f_img o;
o.pos = v.vertex * float4(2, 2, 0, 0) + float4(0, 0, 0, 1);
#if UNITY_UV_STARTS_AT_TOP
o.uv = v.texcoord * float2(1, -1) + float2(0, 1);
#else
o.uv = v.texcoord;
#endif
return o;
}
#if !SHADER_API_GLES // excluding the MRT pass under GLES2
struct CombinerOutput
{
half4 gbuffer0 : SV_Target0;
half4 gbuffer3 : SV_Target1;
};
CombinerOutput frag_gbuffer_combine(v2f_img i)
{
half ao = tex2D(_OcclusionTexture, i.uv).r;
CombinerOutput o;
o.gbuffer0 = half4(0, 0, 0, ao);
o.gbuffer3 = half4((half3)EncodeAO(ao), 0);
return o;
}
#else
fixed4 frag_gbuffer_combine(v2f_img i) : SV_Target0
{
return 0;
}
#endif
// Debug blit
half4 frag_blit_ao(v2f_multitex i) : SV_Target
{
half4 src = tex2D(_MainTex, i.uv0);
half ao = tex2D(_OcclusionTexture, i.uv1).r;
return half4(CombineOcclusion(1, ao), src.a);
}